Debug production without stopping the world (Go)

Steps

1. 1Correlate user report to trace IDs through gateway → service → data layer
2. 2Capture state non-invasively: heap/profile snapshots over process restarts
3. 3Enable dynamic log levels temporarily rather than redeploying
4. 4Test hypotheses against staging replicas of prod data shape
5. 5Timebox live debugging; mitigate via rollback/flag if past window
6. 6Write postmortem linking timeline to evidence artifacts
7. 7Run pprof profiles; check goroutine leaks with goleak
8. 8Enforce vet/lint gates; keep error wrapping with %w intact

Common Pitfalls

• ▲Restarting away evidence mid-diagnosis
• ▲Debugging against sanitized data missing the trigger shape

Commands

$ npx skills add aniruddhaadak80/skills --skill debug-craft-production-debugging-go

$ npx skills add aniruddhaadak80/skills --skill debug-craft-production-debugging-go -g

Tags